This invention relates generally to the field of combined semiconductor and superconductor structures. More specifically, it relates to the field of monolithically integrated high-T.sub.c superconductor/semiconductor structures.
High-temperature superconductors are superconductors that have superconducting transition temperatures (T.sub.c) above about 30 K. These materials are usually mixed oxides of rare earth elements and copper, and are known variously as cuprate superconductors, oxide superconductors, ceramic superconductors, perovskite superconductors, and simply HTS (for high-temperature superconductor) materials. Some examples are La.sub.2-x Sr.sub.x CuO.sub.4 (0&lt;x&lt;0.6), YBa.sub.2 Cu.sub.3 O.sub.7-.delta. (0.1&lt;.delta.&lt;0.6) (YBCO), Tl.sub.2 Ba.sub.2 CaCu.sub.2 O.sub.8 (TBCCO), and Bi.sub.2 Sr.sub.2 CaCu.sub.2 O.sub.8 (BiSCCO). Each of these materials has a variable composition, so the subscripts indicate the approximate midpoint of a range rather than a true stoichiometric ratio. The oxygen content, in particular, is variable, and the oxidation state of the material may affect its superconducting properties; whether or not indicated, the oxygen subscript is assumed to be variable by plus or minus a fraction (usually .delta.) less than unity.
YBCO, TBCCO, and BiSCCO all have phases which have transition temperatures above 90 K, allowing them to operate at 77 K with plenty of engineering margin. This is the temperature at which nitrogen boils, also known as "liquid nitrogen temperature," and is a convenient operating point for many cryogenic systems.
In the seven years since the discovery of high-temperature superconductivity, the field of superconductive electronics has undergone explosive development. Superconducting materials and devices have some unique advantages over conventional materials and semiconductor devices. For example, superconducting interconnects in integrated circuits would lower chip power dissipation while allowing reduced interconnect width and pitch. Superconducting interconnects also would reduce signal pulse degradation due to dispersion, allowing faster integrated circuits (ICs) to be fabricated.
Josephson junction (superconductor) devices have inherent switching speeds that are much faster than semiconductor devices. Josephson junction (JJ) integrated circuits which provide subpicosecond timing resolution have been demonstrated. Entire microprocessors have been successfully built in low temperature superconducting logic; these JJ IC's are two orders of magnitude faster than equivalent gallium arsenide (GaAs) devices and consume two orders of magnitude less power. Superconducting quantum interference devices (SQUIDs) provide magnetic field sensitivity at the quantum limit by exploiting the sensitivity of the phase of the superconducting wavefunction to magnetic fields. SQUIDS, flux-flow transistors (FFTs) with 150 GHz speed, low-loss microwave components, free-standing microstructures and sensitive long-wavelength bolometers have all been demonstrated using HTS materials.
Semiconductor materials are even more widespread, having a longer history. A semiconductor is an insulator at extremely low temperature, and its resistivity drops as temperature increases. Semiconductors can be doped, that is, impurities can be added to change the concentration and sign of charge carriers. By exploiting this property of semiconductors, various junction types can be fabricated and a multitude of electronic devices and circuits can be made using those junctions. This property also accounts, in part, for the ability to make high-density circuits of semiconductors, since it is possible to make several junctions very close together merely by changing the concentration of dopants abruptly.
Operation of semiconductor devices at liquid nitrogen temperature has advantages under certain conditions. A factor of two increase in speed for complementary metal-oxide-semiconductor (CMOS) devices can be obtained without any device design changes merely by operating at low temperatures. This is very significant for ultra-high performance computer systems. Even larger CMOS speed increases can be realized when the devices are optimized for operation at cryogenic temperatures.
While superconducting devices have much higher theoretical limits for speed than semiconductor devices, semiconductor devices perform certain functions better, at least in the current state of the art. For example, while the fastest memory produced is superconductor-based, it has low spatial density. High-density memory cells are difficult to produce in superconducting logic because adjacent cells couple magnetically. Furthermore, superconductor device signal levels are typically millivolts or lower, making such circuits difficult to interface to conventional electronics. Voltage state latches for readouts are difficult to design and implement in high temperature superconductors because HTS junctions are not hysteretic (bistable). On the other hand, semiconductor devices can easily be used to form latches, high-density memories, and level-shifting interface circuits.
As the operating temperatures of superconductors have increased and those of semiconductors have decreased, the suggestion to combine the two has been heard more and more often. The combination of both superconductor and semiconductor devices on the same chip operated at 77 K, for example, would exploit the unique advantages of each technology; the symbiosis leads to performance unattainable by either alone.
Soon after the discovery of HTS, visions of high-current-carrying, zero resistance interconnects for semiconductor circuits, as well as high-speed hybrid circuits combining semiconducting and superconducting logic danced in the heads of many researchers. The realization of these visions, however, has remained elusive. This is due in large part to the chemistry of HTS materials and their high content of copper, an anathema for semiconducting circuits.
There are several major obstacles to the integration of high temperature superconductors with semiconductors, and particularly with silicon (Si) or gallium arsenide (GaAs) devices. The HTS materials are oxides which are very weakly bound chemically, and are therefore easily decomposed (often into elemental copper and other metals) by direct contact with semiconductors. Elemental copper is an "interstitial-substitutional" diffuser in Si, Ge, and GaAs, as well as in other semiconductors, and diffuses through these materials about as fast as Li does, i.e., faster than any other element except hydrogen. At typical processing temperatures, Cu can move centimeters in a few hours' time. Copper forms a deep electronic trap in Si, Ge, and GaAs, and even at part per million levels it destroys the very properties of the semiconductor which allow it to be used to make transistors and integrated circuits.
Several attempts have been made to fabricate HTS circuits or devices directly on a semiconductor material. Although simple HTS films can be grown on buffered Si, they have never been demonstrated to be useful for making Josephson junctions. Fork, et al., (Appl. Phys. Lett., vol. 57, p. 1161, 1990) have reported the growth of YBCO on Si. The thickness of the superconductor layer was stress-limited to less than about 150 nm, and the critical current density at useful thicknesses of about 130 nm was as low as 10.sup.5 A/cm.sup.2 at 77 K. Most useful devices require a critical current density greater than or equal to 10.sup.6 A/cm.sup.2 at 77 K. Cracks appeared in all films thicker than about 50 nm, and the transition temperature of the superconductor was depressed. No devices were fabricated in the superconductor or the semiconductor before or after the growth of the YBCO.
This group has also reported the growth of YBCO on GaAs (Appl. Phys. Lett., vol. 60, p. 1621, 1991). The thickness of the superconductor layer was again stress-limited, but not as severely: layers less than about 100 nm showed no signs of fracture. The critical current density at 77 K was about 10.sup.5 A/cm.sup.2, and the transition temperature of the superconductor was again depressed. No devices were fabricated in the superconductor or the semiconductor before or after the growth of the YBCO. A further problem posed by direct growth on GaAs is its tendency to lose As at elevated temperatures, as noted by Fork. Buffer layers may help limit this out-diffusion if they can be deposited at low enough temperatures, but any loss at all can destroy the functionality of the GaAs devices.
To date, there have been no reports of the integration of functional semiconductor and superconductive devices on a single substrate.